Herschel Calibration Workshop
March 25 – 27, 2013
ZeroPoint Correction for
Herschel-SPIRE Maps via PlanckHFI Cross Calibration
Bernhard Schulz, Luca Conversi
with special thanks to
Guilaine Lagache, Pedro Gomez, Steve Guest, Gabriele Mainetti,
the Planck Team, the SPIRE team, and the HCalSG
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
Overview
•  Zero-point correction:
–  Objective, Plan, Change of Plan
• 
• 
• 
• 
• 
Algorithm for comparison
Color correction
Derivation of beam profiles
HIPE 10 implementation
New HIPE 10 products
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
SPIRE/HFI Cross-Calibration
The Opportunity
The Plan
• 
Herschel-SPIRE 350mm and 500mm
filters have strong overlaps with PlanckHFI 857GHz and 545GHz respectively
• 
• 
Planck-HFI use COBE-FIRAS maps (3%
accuracy) to flux-calibrate their maps in
these two filters in offset and scale.
• 
• 
The COBE-FIRAS calibration for the
overlapping filters is based on two onboard blackbodies operated in space.
• 
The FIRAS beam is 7°, the HFI beam is
5’, the SPIRE beams are 25” and 36”.
Take accurate blackbody calibrated
COBE-FIRAS all-sky maps with 7° beam
and cross calibrate Planck-HFI 545 and
857 GHz maps with ~5’ beams.
Take Planck-HFI 545 and 857 GHz
maps, make sure photometric gains
match, and correct zero-point of SPIRE
maps with 18, 25, 36” beams.
HFI-857
PLW
PMW
HFI-545
PSW
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
Change of Plans
•  In our first comparisons
we found the photometric
gains of the FIRAS based
HFI maps were
considerably higher than
those of the SPIRE maps.
•  Because of this and other
inconsistencies the
Planck team decided to
change the calibration
scheme for the 545 and
857 GHz channels.
•  Change of calibration
scheme for Planck-HFI
545 and 857 GHz maps
–  Derive photometric gain
from observations of
Uranus and Neptune.
–  Derive zero levels from
correlation of HI (21cm)
gas column density and
add CIB mean level.
–  See talk by Guilaine
Lagache (Planck Collaboration
VIII. 2013, In preparation)
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
Map Comparison (Algorithm)
•  Derive a color corrected HFI map at the
reference wavelength of the SPIRE band based
on the flux ratio of both HFI bands.
•  Convolve SPIRE maps assuming an 8’
Gaussian beam.
•  Determine first offset estimate and embed
SPIRE map into HFI map that is set to SPIRE
offset estimate to reduce edge effects.
•  Repeat convolution and comparison.
•  Comparison (for photometric verification)
–  Plot HFI and SPIRE maps pixel by pixel and fit
gain and offset.
–  Determine median and scatter in difference map.
Planck simulated map
SPIRE embedded
+ convolved
•  Correction (for SPIRE pipeline)
–  Determine median and scatter in difference map.
–  Add median difference to SPIRE map
page ‹#›
PACS
response
function
the instrument for
T( ν) is the relative spectralSED
not only
at
the of
reference
ν01 to the real sky spectrum Fwhere
document
describes
aFway( to
ν)perform
ν
sky ( ν). This fixes the flux of thatThis
ref
Herschel Calibration Workshop
and
overlapping
SPIRE
and
HFI
filterbands
tha
frequency of the first filter, but Aalso
for calibration
the secondoffilter.
ii) Then the
flux at thefilters
reference
frequency
cross
two different
photometric
can be
done with ob
March 25 – 27, 2013
To
correct
this
flux
for
the
actual
spectrum
of
th
without
user
intervention.
ν
to the
fiducial
spectrum.
Let be
ν02 of the second filter is transferred
Fstd
ν01asand
area onback
the sky
with
knownreference
relative
This can
seen
a sequenc
N 1 SED.
standard Herschel or1 Planck photometer fluxde
Introduction
1
ν0
ν T1( ν) dνν0 and ν0
fluxdensities
at theFsky
reference
wavelengths
Fstd ν02 be the monochromatic
steps:
i) first a transfer
from 1the fiducial
reference1spectrum2 S = const ν
ν
0
The
Herschel
and
Planck
missions
quote
thecorrection
actual
at
theSED
referen
Fsky
the respective
color
factors
asthe
defined
respectively, and let K1 and Kν0
skyof
spectrum
This of
fixes
thesky
fluxabove.Then
of that
not o
Fsky ( νSED
).
2 be
1 to the real
ν flux densities S at a given reference wave
N 1
from
1 but also for the second filter. ii) Then the flux a
frequency of the first filter,
Fsky ( ν ) T1( ν ) dν
ν0 1
Fref ( ν)
of
the
second
filter
is
transferred
back to the fiducial
reference
spectru
ν0
Fstd ν02 K2
Fsky ν02 2
K
F
ν0
ν
F
ν0
1the2sky
2spectrum
sky
0
Tofluxdensities
correct
thisν0
flux
forreference
actualwavelength
follows
K12monochromatic
=
=
Fstd
ν0
= Fstd
the
at
the
Fstd ν02 be
2
1
ν
ν0F1
Fsky ν01
ν01
Fstd ν01 K1
K2Planck
standard
HerschelFsky
or
f
N 1
sky photometer
with
1respective
F
(
ν0
)
K
=
F
(
ν0
)
K
and
be
the
color
correction
factors
respectively, and let K
K
Fsky
ν0 2 2
ν T2sky
( ν) dν
1 std
actual SED of the sky at the re
Fsky of the
ν
from
0
Color Correction
• 
• 
• 
• 
• 
Objective: From 857 and 545 GHz
maps construct a map as if
with
,
measured with SPIRE.
The HFI and SPIRE maps are
N 1
Fstd ν02 K2
Fsky νν0
K1 Fsky ν02
2
1
defined at the reference
=the
F
ν02 = Fs
so the correction factor from one fluxdensity toν0
other
becomes
Fskyand
( ν ) Tfollows
( ν )Kdν =
2
2
12 K F std
Fsky ν01
Fstd ν01 K1
ν0
wavelengths/frequencies and a
F
(
ν0
)
K
=
F
(
ν0
)
ν
2 skysky 1 with
std
0
where T( ν) is the relative spectral response fun
standard reference spectrum.
becomes
the finally
color correction factor becomes
The actual spectrum is a cold
A cross calibration
of two different photometric
νN 1
galactic dust spectrum
ν
spectral
respons
( νN) 1is the relative
so the correction factor from
fluxdensity
to the other
becomes
T1where
( ν)one T
area
on
the
sky
with
known
relative
SED.
This c
ν01
dν
Fsky( ν) T2 ( ν) dν
characterized by a temperature.
ν
A cross νcalibration
of two different photom
0from the
ν 0 a transfer
steps: i) first
fiducial referen
Color correction factors are
area on the sky with known relative SED.
K12 =
νN 1
tabulated for a range of
νN 1
( ν) spectrum
This
fixes
ν01 to the realTsky
F from
( ν).the
i) first
a transfer
fiducial
re
2steps:
temperatures.
dν
ν02 3
Fsky( ν) sky
T1 ( ν) dν
νν0 to the real sky spectrum F
). This
2 h nu
1
filter,
but also for sky
the( νseco
ν0
Using the Planck flux
law
B( nu T)frequency νof the first
Ratios of monochromatic
0
h nu
frequency
of the
first filter, but
also for
filter is
transferred
back
to the
th
ν02 of the second
densities in the HFI map are
2
kb T
The
sky
SED
is
of
the
second
filter
is
transferred
back
ν0
c e
1
2betwen HFI and SPIRE
Color Correction
Factors
tabulated for the same list of
monochromatic
fluxdensities
Fstd ν02 be the
3 fluxdensa
Fstd ν02 be the monochromatic
1.8
2the
h nu
we
will
calculate
the
color
corrrection
factors
for
practical
cas
nu
temperatures. we define the relative sky SED as In theF following
Using
B( nu T)
with
( nu T)the Planck
B( nu T)law
and
be
the
rest
respectively,
and
let
K
K
fluxessky
measured
with the two
highest
frequency
filters
of
Planck-HFI
into
fluxes
h
nu
and
be
the
respecti
respectively,
and
let
K
K
GHz
1
2
1 filters. It 2is assumed that the com
correspond to the overlapping Herschel-SPIRE
2
kb T
from
c e
1
corresponding
skybins
of regridded
maps that
have
from
An alternative to the SED based on the Planck
law could
be a simple
power spectrum
of the
formbeen convolved to the same
must be larger or at least equal to the instrument
with K
the smaller
Fstd ν0
Fsky spatial
ν02 resolut
1.8
2 2
nu
b
=
we
define
the
relative
sky
SED
as
F
(
nu
T
)
B
(
nu
T
)
nu
Fstd ν02 K
ν02 F ν0 GHzHl fo
sky K
Hs2 F Fν0
sky
Fsky_p( nu b ) M
Jy
1
sky 1 follows
std
1
..\bandpass_857_v1.0.da
GHz...\SpirePhotRSRF.txt
and the flux ratio is
page ‹#›
10
=
K
Fsky
ν01be a simple power spe
F
ν0
An alternative to the SED
law could
1 Planck
std based1 on the
M
reverse ( M )
b
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
Neptune Fine Scan Maps
PSW
PMW
PLW
•  With background sources.
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
Fitted Background
Ecliptic
coordinate grid
Direction of tilt and strength of gradient are compatible
with using a Zodiacal light model by Reach et al. (1996)
and extrapolating it to submm wavelengths.
•  Solution 1:
•  Fit and subtract a
warped background
plane (implemented).
•  Solution 2:
•  Obtain “shadow”
observation and
subtract (next version).
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
Neptune Fine Scan Maps
PSW
r=1000’’
PMW
PLW
r=700’’
•  Maps after removal of background sources.
•  Coverage beyond 700’’ radius is low and reliability is questionable.
page ‹#›
PACS
200
400
600
radius [arcsec]
800
0
1000
Herschel Calibration Workshop
March 25 – 27, 2013
PMW radial flux average [Jy] (tilt-sub)
PSW radial flux average [Jy] (tilt-sub)
100
Sector 1
Sector 2
Sector 3
10-2
10-4
200
400
600
radius [arcsec]
800
1000
0
100
200
400
600
radius [arcsec]
800
1000
100
PLW radial flux average [Jy] (tilt-sub)
0
Sector 1
Sector 2
Sector 3
10-2
10-4
Sector 1
Sector 2
Sector 3
10-2
10-4
After Source and Background Removal
400
600
radius [arcsec]
800
PMW
radial
average
(tilt-sub)
PMW
radial
fluxflux
average
[Jy][Jy]
(tilt-sub)
0
Sector 1
Sector 2
Sector 3
-1•10-4
-1•10-4 0
0
200
200
400
400
600
radius [arcsec]
600
radius [arcsec]
800
800
1000
1000
1•10-4-4
1•10
5•10-5-5
5•10
Background [Jy/beam] = :-8.0E-06
0
0
After background subtraction of [Jy/beam] = :-8.0E-06
-5•10-5-5
-5•10
-1•10-4-4
-1•10
0
200
200
400
600
400 Inner radius600
[arcsec]
radius [arcsec]
800
800
1000
1000
1200
462’’2
-5
5•10
460
-1•10-4
0
0
200
200
400
400
600
800
600
800
Integration radius [arcsec]
Inner radius [arcsec]
1000
1000
-5•10-5
-5•10-5
1000
0
Sector 1
Sector 2
Sector 3
-4
-1•10
-1•10-4 0
0
200
200
400
400
600
radius [arcsec]
600
radius [arcsec]
800
800
1000
1000
200
400
600
radius [arcsec]
800
1000
5•10-5-5
5•10
Background [Jy/beam] = :-1.8E-05
0
0
After background subtraction of [Jy/beam] = :-1.8E-05
-5•10-5-5
-5•10
-1•10-4-4
-1•10
0
0
200
200
400
400
600
600
Inner radius [arcsec]
radius [arcsec]
800
800
1000
1000
1200
825’’2
830
-5
795’’2
8100
Omega(Neptune) [arcsec^2] = : 794.8 +/- 0.3
After background subtraction of [Jy/beam] = :-1.8E-05
200
200
400
600
800
400 Integration radius
600 [arcsec] 800
Inner radius [arcsec]
1000
1000
Background [Jy/beam] = :-2.0E-05
5•10-5
Background [Jy/beam] = :-2.0E-05
1200
1200
3
0
-5•10-5
2
Sector 1
Sector 2
Sector 3
-5•10-5
-1•10-4
0
200
400
0
200
400
600
radius [arcsec]
600
radius [arcsec]
800
1000
800
1000
1•10-4-4
1•10
5•10-5-5
5•10
Background [Jy/beam] = :-2.0E-05
0
0
After background subtraction of [Jy/beam] = :-2.0E-05
-5
-5
-5•10
-5•10
-1•10-4-4
-1•10 0
0
200
200
400
600
400 Inner radius600
[arcsec]
radius [arcsec]
800
800
1000
1000
1200
1690’’2
1665’’2
1•10-4
1690
5•10-5
1680
0
1670
Omega(Neptune) [arcsec^2] = :1665.4 +/- 0.2
After background subtraction of [Jy/beam] = :-2.0E-05
-1•10-40
0
200
200
400
600
800
400 Integration 600
radius [arcsec] 800
Inner radius [arcsec]
1000
1000
1200
1200
Removed sources and warped plane improves consistency between three beam sectors.
Background estimates similar to previous estimates before source removal.
Integration out to 600’’ results in 450’’2, 795’’2, 1665’’2 respectively for PSW, PMW, PLW.
Integration out to 1000’’ results in 462’’2, 825’’2, 1690’’2 respectively for PSW, PMW, PLW.
Integrated solid angles smaller again.
465
460
455
Omega(Neptune) [arcsec^2] = : 449.5 +/- 0.1
450
0
200
400
600
800
Integration radius [arcsec]
1000
1200
830
820
810
Omega(Neptune) [arcsec^2] = : 794.8 +/- 0.3
800
790
0
200
400
600
800
Integration radius [arcsec]
1000
1200
1
0
1660-5
-5•10
790
-1•10-4
0
0
1•10-4
5•10-5
-1•10-4
1•10-4-4
1•10
800
-5•10-5
1200
1200
800
0
0
5•10
820
After background
subtraction
= :-8.0E-06
Omega(Neptune)
[arcsec^2]
= : 449.5of
+/-[Jy/beam]
0.1
-5
600
radius [arcsec]
Background [Jy/beam] = :-1.8E-05
Background [Jy/beam] = :-1.8E-05
1•10-4
450’’2
0
455
-5•10
450
PSW Integrated Solid Angle [arcsec^2])
0
PMW
Average
background
outside
radius [Jy]
PMW
radial flux
average [Jy]
(tilt-sub)
-5•10-5
400
1•10-4
5•10-5
5•10-5
PMW Average background outside radius [Jy]
PMW Integrated Solid Angle [arcsec^2])
-5•10-5
200
1•10-4
1•10-4
PMW Integrated Solid Angle [arcsec^2])
radial
average
(tilt-sub)
PSWPSW
radial
fluxflux
average
[Jy] [Jy]
(tilt-sub)
0
1•10-4
465
• 
• 
• 
• 
• 
Background [Jy/beam] = :-8.0E-06
Background [Jy/beam] = :-8.0E-06
5•10-5
PSW Average background outside radius [Jy]
PSW Integrated Solid Angle [arcsec^2])
PSW
background
outside
radius [Jy]
PSWAverage
radial flux
average [Jy]
(tilt-sub)
1•10-4
10-6
10-8
0
1•10-4
5•10-5
10-8
1000
flux average
[Jy] (tilt-sub)
PLWPLW
radialradial
flux average
[Jy] (tilt-sub)
200
PLW
background
outside
radius [Jy]
PLWAverage
radial flux
average [Jy]
(tilt-sub)
0
PLW Average background outside radius [Jy]
PLW Integrated Solid Angle [arcsec^2])
10-8
10-6
PLW Integrated Solid Angle [arcsec^2])
10-6
1690
1680
Omega(Neptune) [arcsec^2] = :1665.4 +/- 0.2
1670
1660
0
200
400
600
800
Integration radius [arcsec]
1000
1200
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
SPIRE Extended Source calibration
• 
• 
• 
• 
• 
• 
Re-built fine scan map of Neptune with 1
arcsec pixel size.
Removed sky background structure (tilt
and galaxies)
Integrated solid angle to 600” radius.
Estimated uncertainty is +/- 4%.
Recently (Nov 2012) obtained “shadow”
observation of the same region should
improve accuracy further.
Monochromatic extended source fluxes
are quoted assuming the standard
ν*Fν=const. reference spectrum.
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
HIPE 10 Implementation
•  New Healpix Library to read HFI 545 GHz and HFI 857
GHz maps.
•  Interpolator for irregularly distributed data points.
•  HIPE task “zeroPointCorrection()”
•  Input parameters:
–  Level 2 context with three extended source maps
–  Two HFI Healpix maps at 545 GHz and 857 GHz.
–  A table with pre-calculated relative color correction factors
based on relative HFI flux ratios and a simple dust model.
–  Photometric gains (1.0) SPIRE/HFI for 545 GHz and 857
GHz.
–  SPIRE photometer solid angles (hard coded 465, 822,
1768”^2)
–  Effective Gaussian FWHM of HFI maps (8’’)
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
Comparison in Detail I
PMW or PLW
Convolved
SPIRE map
Solid Angle from
Beam Profile
Expand
Map
Convert to MJy/sr
multiply by K4ext/K4pnt
divide by Solid Angle
Convolve
Kernel is convolution of
8’ Gaussian and
HFI Beam model
(Smaller ~3.5’ Gaussian)
SPIRE
All sky Healpix maps
HFI-545
Color Correction
HFI-857
Ratio
map
Color corrected
HFI map
HFI-545 to PLW
HFI-857 to PMW
Ratio HFI-857/HFI-545
HFI
page ‹#›
Color correct HFI map to SPIRE filter bands PSW
or PLW based on fit of Grey-Body dust model with
β=1.8 to HFI fluxes.
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
Comparison in Detail II
Determine
median
difference.
Fit Offset
and Scale.
Compare
within mask
Flux SPIRE Map
Convolved
SPIRE map
Color corrected
HFI map
Embed
SPIRE map
into median
adjusted HFI
map
Embedded
SPIRE map
Flux HFI Map
Mask where
edge effects <0.1%
1st Scale
1st Offset
1st Median Offset
Embedding reduces edge effects in
the subsequent SPIRE map
convolution
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
Comparison in Detail III
Embedded
SPIRE map
Pipeline Output
Convolved
SPIRE map
Color corrected
HFI map
Compare
within mask
Flux SPIRE Map
Convolve
Determine
median
difference.
Fit Offset
and Scale.
SPIRE map
with corrected
zero-point
using median
difference
Flux HFI Map
2nd Scale
2nd Offset
2nd Median Offset
Mask for
entire
SPIRE map
Median and
standard
deviation of
the difference
map in FITS
header
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
Map Correlations
First Comparison
Second Comparison
(embedded map)
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
New last segment of SPIRE standard product generation
timelines with
sky positions
psp product
map product
diagnostic product
psrcPxWdiag
destriper
Level 2
Level 1
baseline
corrected
timelines
insert destriper map
psrcPxW
Naïve map
ssoPxW
ssoMotionCorrection
applyRelativeGains
destriper
extdPxWdiag
zeroPointCorrection
Planck HFI
Maps
Gain Factors
extdPxW
Colour Correction
Factors
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
New Level 2 Products
•  Point source maps [Jy/beam]
renamed to psrcP?W.
•  Associated diagnostic products
named psrcP?Wdiag.
•  (for SSOs) Proper motion
corrected maps ssoP?W.
•  Extended source maps [MJy/sr]
with FIRAS/HFI absolute sky
background levels extdP?W.
•  Associated diagnostic products
named extdP?Wdiag.
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
Planck-HFI Maps
In March 2011 the Planck
team kindly agreed to make a
confidential release of HFI 545
and 857 GHz maps before the
official release date to enable
zero-point correction of SPIRE
maps in the data reduction
pipeline at the Herschel
Science Center only.
• 
Subsequently four different internal releases for the HFI 545 GHz and HFI
857 GHz filter bands were received and evaluated
–  (DR2, DX7, DX9, re-calibrated DX9).
• 
• 
• 
Relative photometric gains, consistency and beam profile verifications were
conducted.
HFI photometry of planets was compared with planet flux models.
Regular telecons/meetings of the HFI/SPIRE cross calibration team were
held to clarify matters, agree on policies, and to provide feedback to the
Planck project.
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
Gain Verification
•  Verification of the photometric gain is a prerequisite for the
zero-point correction of SPIRE photometer maps.
•  Select large SPIRE photometer maps, re-reduce to Level 1
and construct de-striped maps.
•  Exclude smaller fields, fields with low flux dynamic range,
fields with strong saturated sources close to the edge.
•  Run comparison procedure.
–  There are one IDL and two HIPE implementations for cross
checking.
•  Determined ratio of photometric gains for SPIRE and HFI
maps from pixel to pixel plots of SPIRE and HFI data.
•  The scatter of the difference maps is at a few per cent of the
respective background flux level.
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
Map Flux Ranges
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
Gain Verification
• 
• 
A total of 119 large SPIRE maps were
included in the comparison.
We plot the flux ratios PMW/HFI-857
(top) and PLW/HFI-545 (bottom).
• 
The median ratios are 0.949 and 0.951
with standard deviations 0.009 and 0.011
for PMW/HFI-857 and PLW/HFI-545
respectively.
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
Gain Verification
•  The HFI fluxes are about 5% brighter than
the SPIRE fluxes in both channels.
•  A similar comparison of several maps by the
Planck team yields an average of 2% and 7%
brighter HFI fluxes at 545 and 857 GHz
(Planck Collaboration VIII. 2013, In
preparation).
•  Although the correspondence between the
HFI and SPIRE calibrations is very good, the
relative difference between the gains both
teams found is significant and we are looking
for the reason.
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
HIPE and IDL Results
HIPE and IDL gains agree well, on average within 0.4% and 1.7% for PMW and PLW.
page ‹#›
PACS
Herschel Calibration Workshop
March 25 – 27, 2013
Summary
•  The new zero point calibration implementation into HIPE
10 is complete and operational.
• 
• 
• 
• 
• 
• 
• 
The new maps fit into the new Level 2 product structure adopted by SPIRE
for HIPE 10.
The new Level 2 product structure contains zero-point corrected SPIRE
maps in [MJy/sr] and keywords specifying the median and the standard
deviation of the difference map.
New solid angles for the conversion to extended source fluxes have been
determined to an accuracy of 4%.
New “shadow” observations will improve the accuracy further and are being
analyzed.
The current HFI maps are photometrically consistent with the SPIRE maps
within the quoted 10% absolute photometric accuracy of the HFI maps.
The HFI photometry seems to be biased towards slightly higher fluxes and
both teams find different photometric ratios.
A more refined analysis is ongoing.
page ‹#›
PACS